PT  - JOURNAL ARTICLE
AU  - Emilie Prang Nielsen
AU  - Søren Andersen
AU  - Christian Lehn Brand
AU  - Susanne Ditlevsen
TI  - Applying Historical Data in a Nonlinear Mixed-Effects Model Can Reduce the Number of Control Rats Required for Calculation of The Relative Potency of Insulin Analogues
AID  - 10.1101/2021.09.10.459831
DP  - 2021 Jan 01
TA  - bioRxiv
PG  - 2021.09.10.459831
4099  - http://biorxiv.org/content/early/2021/09/10/2021.09.10.459831.short
4100  - http://biorxiv.org/content/early/2021/09/10/2021.09.10.459831.full
AB  - This paper examines how to reduce the number of control animals in preclinical hyperinsulemic glucose clamp studies if we make use of information on historical studies. A data set consisting of 59 studies in rats to investigate new insulin analogues for diabetics, collected in the years 2000 to 2015, is analysed. A simulation experiment is performed based on a carefully built nonlinear mixed-effects model including historical information, comparing results (for the relative log-potency) with the standard approach ignoring previous studies. We find that by including historical information in the form of the mixed-effects model proposed, we can to remove between 23% and 51% of the control rats in the two studies looked closely upon to get the same level of precision on the relative log-potency as in the standard analysis. How to incorporate the historical information in the form of the mixed-effects model is discussed, where both a meta approach as well as a Bayesian approach are suggested. The conclusions are similar for the two approaches, and therefore, we conclude that the inclusion of historical information is beneficial in regard to using fewer control rats.Competing Interest StatementI have read the journal's policy and the authors of this manuscript have the following competing interests: Søren Andersen and Christian Lehn Brand are employed by Novo Nordisk and hold shares in the company.